Author + information
- Received October 29, 2012
- Revision received February 23, 2013
- Accepted March 13, 2013
- Published online September 24, 2013.
- Bonnie Ky, MD, MSCE∗,†∗ (, )
- Benjamin French, PhD∗,†,
- Abigail May Khan, MD∗,
- Ted Plappert, CVT∗,
- Andrew Wang, BA‡,
- Julio A. Chirinos, MD∗,
- James C. Fang, MD§,
- Nancy K. Sweitzer, MD, PhD‖,
- Barry A. Borlaug, MD¶,
- David A. Kass, MD#,
- Martin St. John Sutton, MBBS∗ and
- Thomas P. Cappola, MD, ScM∗
- ∗Penn Cardiovascular Institute, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
- †Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
- ‡Robert Wood Johnson Medical School, Piscataway, New Jersey
- §Division of Cardiology, University of Utah, Salt Lake City, Utah
- ‖Cardiovascular Medicine, University of Wisconsin, Madison, Wisconsin
- ¶Mayo Clinic College of Medicine, Rochester, Minnesota
- #Division of Cardiology, Department of Medicine, Johns Hopkins Medical Institutions, Baltimore, Maryland
- ↵∗Reprint requests and correspondence:
Dr. Bonnie Ky, Department of Cardiovascular Medicine, University of Pennsylvania, 3400 Civic Center Boulevard, TRC 11-05, Philadelphia, Pennsylvania 19104.
Objectives The objective of this study was to compare the physiological determinants of ejection fraction (EF)—ventricular size, contractile function, and ventricular-arterial (VA) interaction—and their associations with clinical outcomes in chronic heart failure (HF).
Background EF is a potent predictor of HF outcomes, but represents a complex summary measure that integrates several components including left ventricular size, contractile function, and VA coupling. The relative importance of each of these parameters in determining prognosis is unknown.
Methods In 466 participants with chronic systolic HF, we derived quantitative echocardiographic measures of EF: cardiac size (end-diastolic volume [EDV]); contractile function (the end-systolic pressure volume relationship slope [Eessb] and intercept [V0]); and VA coupling (arterial elastance [Ea]/Eessb). We determined the association between these parameters and the following adverse outcomes: 1) the combined endpoint of death, cardiac transplantation, or ventricular assist device (VAD) placement; and 2) cardiac hospitalization.
Results Over a median follow-up of 3.4 years, there were 76 deaths, 52 transplantations, 14 VAD placements, and 684 cardiac hospitalizations. EF was independently associated with death, transplantation, and VAD placement (adjusted hazard ratio [HR]: 3.0; 95% confidence interval [CI]: 1.8 to 5.0 comparing third and first tertiles), as were EDV (HR: 2.6; 95% CI: 1.5 to 4.2); V0 (HR: 3.6; 95% CI: 2.1 to 6.1); and Ea/Eessb (HR: 2.1; 95% CI: 1.3 to 3.3). EDV, V0, and Ea/Eessb were also associated with risk of cardiac hospitalization. Eessb was not significantly associated with any adverse outcomes in adjusted analyses.
Conclusions Left ventricular size, V0, and VA coupling are associated with prognosis in systolic HF, but end-systolic elastance (Eessb) is not. Assessment of VA coupling via Ea/Eessb is an additional noninvasively derived metric that can be used to gauge prognosis in human HF.
Accurate assessment of cardiac performance is critical in chronic heart failure (HF) to gauge prognosis and to assess response to therapeutic interventions. Among patients who have HF due to systolic dysfunction, left ventricular (LV) ejection fraction (EF) is the most widely used quantitative measure to characterize cardiac function and is a potent predictor of clinical outcomes (1). Although EF can be simply ascertained from volumetric measures, it is a complex summary measure that integrates several underlying physiological components, including ventricular size, contractile function, and afterload. The gold standard for assessing each of these components is invasive pressure-volume analysis, but validated noninvasive approaches have also been developed (Fig. 1) (2,3). Left ventricular (LV) end-diastolic volume (EDV) is an index of LV size and quantifies the degree of cardiac remodeling. The end-systolic pressure volume relationship (ESPVR) provides a load-independent measure of contractile function. The ESPVR is typically assumed to be linear and is therefore defined by a slope and an intercept (4). Although many studies focus on the slope alone, both the slope (end-systolic elastance [Ees]) and the intercept (V0) are required to describe the contractile state of the left ventricle. Ees quantifies ventricular elastance (stiffness) at end-systole, and V0 is a measure of ventricular volume at a theoretical end systolic pressure of 0 mm Hg. Because V0 is an extrapolated value obtained at a nonphysiological pressure, the ventricular end-systolic volume (ESV) at a systolic pressure of 100 mm Hg (V100) is also often described. Effective arterial elastance (Ea), the negative slope joining the end-systolic pressure-volume point to the point on the volume axis at end-diastole, provides an integrative measure of arterial load. The net interaction of the ventricular and arterial systems is indexed by the ratio Ea/Ees, which strongly influences cardiovascular performance and efficiency (4–6). Despite noninvasive methods to assess these parameters, their relative importance in determining HF prognosis is unknown.
We sought to compare the impact of the component parts of EF (cardiac size, contractile function, and ventricular-arterial [VA] interaction) on prognosis in chronic HF. A secondary objective of this study was to assess the feasibility of ascertaining noninvasively derived, single-beat measures of ventricular and arterial elastance in a large cohort of patients with HF and reduced EF. We performed rigorous quantitative 2-dimensional echocardiography analyses in a subcohort of 466 patients from the Penn Heart Failure Study, an ambulatory population composed of participants with primarily systolic dysfunction. We determined the relationships between EF, ventricular size (EDV), contractile function (Eessb and V0), and VA coupling (Ea/Eessb) and the following clinical outcomes: 1) the combined endpoint of death, cardiac transplantation, and ventricular assist device (VAD) placement; and 2) cardiac hospitalization.
The Penn Heart Failure Study is a prospective cohort study of outpatients with primarily chronic systolic HF recruited from referral centers at the University of Pennsylvania (Philadelphia, Pennsylvania), University of Wisconsin (Madison, Wisconsin), and Case Western (Cleveland, Ohio) (7–9). This substudy consisted of subjects recruited from the University of Pennsylvania representative of the clinical site. The primary inclusion criterion was a clinical diagnosis of HF. Participants were excluded if they had a noncardiac condition resulting in an expected mortality of <6 months, as judged by the treating physician, or if they were unable or unwilling to provide informed consent.
At time of study entry, detailed clinical data were obtained using a standardized questionnaire administered to the patient and treating physician, with verification by medical records. Two-dimensional transthoracic echocardiography was performed in all patients at an Intersocietal Commission for the Accreditation of Echocardiography Laboratories (ICAEL)-accredited laboratory, typically within 60 days of study entry.
Follow-up events including all-cause mortality, cardiac transplantation, VAD placement, and any cardiac hospitalizations were prospectively ascertained every 6 months by direct patient contact and verified by death certificates, medical records, and contact with patients’ family members by dedicated research personnel.
All participants provided written, informed consent, and the Penn Heart Failure Study protocol was approved by the institutional review board.
Digital echocardiograms were analyzed on TomTec computer workstations (TomTec Imaging Systems, Unterschleissheim, Germany). Apical 4-chamber LV EDV and ESV were obtained using Simpson’s method of disks as recommended by the American Society of Echocardiography (3). LV volumes and mass were indexed to body surface area, which was determined using the Dubois formula: (0.20247 × height (m)0.725 × weight (kg)0.425).
ESP was estimated as 0.90 × systolic pressure, obtained by manual blood pressure cuff measurement (2). Stroke volume (SV) was measured from the LV outflow tract (LVOT) diameter and the pulse wave Doppler signal (2). EF was calculated from 2-dimensionally derived LVOT area and the SV (10). Effective Ea was defined as the ratio of ESP/SV (10). Ees was determined using a modified single-beat algorithm described by Chen et al. (2) using arm cuff pressures, echocardiography-derived Doppler SVs, and several timing intervals (isovolumic contraction time, pre-ejection period, ejection time, total systolic period) and is denoted as Eessb (2). Ventricular-vascular coupling was estimated by the Ea/Eessb ratio. Additional indices of the ESPVR and LV size, V0 and V100, were also estimated from Eessb, ESV, and ESP.
Approximately 5% of the patients had LV apical tracings of limited image quality that precluded adequate quantitation. Doppler tracings, timing intervals, and LVOT diameter could not be obtained in ∼5% to 8% of participants secondary to limitations in image quality. After exclusion of these individuals, we had a complete dataset of 466 participants.
For EDVs, the intraobserver coefficients of variation (CVs) for this measurement in our core lab are 4.5% to 6.3%. The intra- and interobserver CVs for Eessb are 8.2% and 9.8% and for Ea are 7.9% and 8.6%, respectively. The intra- and interobserver CVs for SV based on the LVOT method are 7.9% and 8.6%, and the intraobserver CV for EF is 10%.
N-terminal pro–B-type natriuretic peptide (NT-proBNP) was measured from banked plasma obtained at the time of study entry by a standard electrochemiluminesence immunoassay (Elecsys proBNP, Roche Diagnostics, Indianapolis, Indiana), as previously described (11). The assay range was 20 to 5,000 pg/ml. The intra-assay and interassay CVs were 2.9% and 6.1%, respectively.
Baseline characteristics were summarized for all participants using standard descriptive statistics. Echocardiographic parameters were evaluated according to the New York Heart Association (NYHA) functional class using the Kruskal-Wallis rank sum test, and the relationships between echocardiographic parameters and NT-proBNP were assessed using Spearman rank correlations. Cox regression models were used to determine the association between EF, EDV, Ea, Eessb, Ea/Eessb, and V0 and V100, and risk of the combined outcome of all-cause death, cardiac transplantation, or VAD placement. To facilitate the comparison of associations across echocardiographic indices, each index was categorized according to tertiles of its distribution. Adjustment variables were selected based on clinical rationale and included age, sex, race, height, weight, heart rate, HF etiology, and medication use (angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, aldosterone antagonists, and beta-blockers). Age exhibited nonproportional hazards and was adjusted for using a time-varying covariate, which was obtained by interacting age with the natural log of time. A priori, adjustment for cardiac resynchronization or defibrillator therapy was not made due to confounding by indication between presence/absence of device therapy and degree of ventricular dysfunction. Recurrent event models were used to determine the association between echocardiographic indices and risk of incident cardiac hospitalization. A joint frailty model was used to accommodate informative censoring by all-cause death, cardiac transplantation, or VAD placement (12). Model complexity precluded adjustment for multiple confounders. Finally, as the primary objective of our analyses was to compare the strength of association across echocardiographic metrics and not the independent associations or incremental value of these measures, we did not place multiple echocardiographic parameters in a single multivariable model, particularly given that many of these parameters are derived from one another, physiologically interdependent, and highly correlated (Online Fig. 1).
All analyses were completed using R 2.15 (R Development Core Team, Vienna, Austria), including the frailty pack extension package (13), and Stata version 12 (StataCorp., College Station, Texas).
Baseline characteristics of the 466 participants are detailed in Table 1. Across the cohort, the median age was 56 years and the majority of the participants were male (63%) and white (73%). The median EF was 27% (interquartile range [IQR]: 17% to 39%), EDV was 115 ml/m2 (IQR: 87 to 150 ml/m2), ESV was 84 ml/m2 (IQR: 58 to 117 ml/m2), and V0 and V100 were similarly increased at 43 ml (IQR: −2.6 to 113 ml) and 167 ml (IQR: 109 to 238 ml), respectively. Across the cohort, the median Ea was substantially elevated at 1.66 mm Hg/ml (IQR: 1.33 to 2.17 mm Hg/ml), and Eessb was markedly reduced at 0.89 mm Hg/ml (IQR: 0.69 to 1.18 mm Hg/ml) (14). The VA coupling ratio (Ea/Eessb) was near 2.0 (median: 1.92; IQR: 1.43 to 2.54), more than twice the value observed in similarly aged, nonfailing subjects (range 0.62 to 0.69) (14).
Associations between echocardiographic parameters and HF severity
We then determined the relationships between these echocardiography-derived measures of size, contractile function, and VA coupling and 2 commonly accepted measures of HF severity: NYHA functional class and serum NT-proBNP levels. As expected, EF and EDV were strongly associated with NYHA functional class (Table 2). Ea and Ea/Eessb were also strongly associated NYHA functional Class, but Eessb was not. V0 and V100, measures indicative of the ESPVR and LV size, were also strongly associated with NYHA functional class. NT-proBNP was significantly correlated with EF (ρ = −0.48; p < 0.001), EDV (ρ = 0.35; p < 0.001), Ea (ρ = 0.28; p < 0.001), Ea/Eessb (ρ = 0.31; p < 0.001), V0 (ρ = 0.31; p < 0.001), and V100 (ρ = 0.40; p < 0.001). NT-proBNP was not correlated with Eessb (ρ = −0.07; p = 0.21).
These findings suggest that there are significant relationships between clinical and laboratory measures reflective of HF severity and echocardiographic parameters reflective of cardiac size (EDV), function and remodeling (V0) arterial load (Ea), and coupling (Ea/Eessb). Interestingly, end-systolic elastance as assessed by Eessb alone was not related to these measures of HF severity.
Associations between echocardiographic parameters and death, cardiac transplantation, and VAD
Over a median follow-up of 3.4 years, there were 141 events, including 76 deaths, 52 transplantations, and 14 VAD placements. EF was strongly associated with these outcomes, with a greater than 3-fold risk in unadjusted and adjusted models when comparing the third and first tertiles (Table 3, Fig. 2). EDV was also significantly associated with adverse outcomes, with a hazard ratio (HR) of 2.7 (95% confidence interval [CI]: 1.8 to 4.0; p < 0.001, third tertile vs. first tertile) in unadjusted models. After adjustment for age, sex, race, height, weight, heart rate, HF etiology, and medication use including angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, aldosterone antagonists, and beta-blockers, this risk was similar (HR: 2.6, 95% CI: 1.5 to 4.2, p < 0.001; third tertile vs. first EDV tertile). For Ea, in unadjusted models comparing the third and first tertiles, those patients with an Ea ≥2.00 mm Hg/ml had a 1.7-fold increased risk of adverse outcomes compared with those patients with an Ea <1.44 mm Hg/ml (95% CI: 1.2 to 2.6, p = 0.008) (Table 3, Fig. 2). After adjustment for potential confounders, this association remained significant (HR: 1.7, 95% CI: 1.1 to 2.6, p = 0.028). In contrast, Eessb alone was not significantly associated with adverse outcomes in unadjusted or adjusted models (p > 0.05 for both). Both V0 and V100 were strongly associated with outcomes in unadjusted (HR: 4.0, 95% CI: 2.5 to 6.4, p < 0.001 for V0 and HR: 3.5, 95% CI: 2.2 to 5.3 for V100, p < 0.001, comparing third and first tertiles) and fully adjusted (HR: 3.6, 95% CI: 2.1 to 6.1, p < 0.001 for V0 and HR: 2.9, 95% CI: 1.7 to 4.8, p < 0.001 for V100) models. When both Eessb and V0 were placed in a single model, V0 retained its significance in both unadjusted (HR: 4.2, 95% CI: 2.6 to 6.7, p < 0.001) and adjusted (HR: 3.8, 95% CI: 2.2 to 6.4, p < 0.001) models, whereas Eessb remained nonsignificant (p > 0.05). VA coupling (Ea/Eessb) was also significantly associated with adverse clinical outcomes in unadjusted (HR: 2.6, 95% CI: 1.7 to 4.0, p < 0.001) and fully adjusted (HR: 2.1, 95% CI: 1.3 to 3.3, p = 0.002) models, with a greater than 2-fold increased risk of adverse outcomes in patients in the highest tertile compared with the lowest. Overall, these findings suggest that EDV, V0, and VA interaction (Ea/Eessb) are the physiological determinants of EF that are most strongly associated with prognosis in humans with systolic dysfunction.
Associations between echocardiographic parameters and cardiac hospitalization
Over a median follow-up of 3.4 years, 272 participants (58% of 466) experienced 684 cardiac hospitalizations before death, transplantation, or VAD placement: 115 (25%) had 1; 62 (13%) had 2; and 95 (20%) had ≥3. EF was strongly associated with an increased risk of cardiac hospitalization (HR: 2.4, CI: 1.8 to 3.2, p < 0.001 comparing those patients with an EF <19% with those with an EF ≥35%) (Table 4). EDV was also significantly associated with cardiac hospitalization, with a similar effect size (HR: 2.1, 95% CI: 1.6 to 2.9, p < 0.001). Individually, Ea was not significantly associated with cardiac hospitalization, and the association of Eessb was of marginal significance in unadjusted models. However, V0 and V100 were significantly associated with cardiac hospitalization (HR: 2.3, 95% CI: 1.8 to 3.1, p < 0.001 for V0 and HR: 2.5, 95% CI: 1.9 to 3.3, p < 0.001 for V100 comparing third and first tertiles). V0 remained significant when Eessb and V0 were placed in a single model (HR: 2.5, 95% CI: 1.9 to 3.3, p < 0.001), whereas Eessb was not. Ea/Eessb was associated with a greater than 2-fold increased risk when comparing the third and first tertiles (Ea/Eessb ≥2.34 vs. 1.60; HR: 2.1, 95% CI: 1.6 to 2.9, p < 0.001). Overall, our findings were consistent across both sets of clinical outcomes: death, cardiac transplantation, and VAD placement and cardiac hospitalization.
EF is an easily obtained and widely used parameter that has unequivocal prognostic importance in human HF. Our primary objective in this study was to compare the strength of the associations of the physiological determinants of EF–LV size (EDV), contractile function (Eessb and V0), and VA coupling (Ea/Eessb) with adverse outcome. In a large cohort of 466 patients with HF and reduced EF, we found that EDV, V0, and VA coupling (Ea/Eessb) were strongly associated with NYHA functional class, increase in natriuretic peptides, and adverse clinical outcomes. Surprisingly, Eessb, a noninvasive measure of the slope of the ESPVR (Fig. 1), showed no significant associations with adverse outcomes. Our results suggest that in the setting of chronic HF, the extent of ventricular remodeling and VA coupling are important indicators of prognosis, whereas Ees is not.
These findings reinforce the well-known hemodynamic importance of aggressive afterload reduction in HF with reduced EF and demonstrate that the pathophysiological significance of chamber elastance alone is less relevant than the matching of ventricular and arterial elastance. For maximal cardiac work, power, and efficiency, the coupling ratio of Ea/Ees typically resides between 0.5 and 1.2 (15,16). In failing hearts, this ratio increases as cardiac function declines and arterial load increases to maintain systolic pressure (17–20). At excessively high ratios, ventricular-vascular matching is significantly compromised, leading to inefficient and ineffective contraction (21).
As expected, we found that Ees, as assessed by Eessb, was low in chronic HF patients (22), but the absence of any association between Eessb and adverse outcomes was initially counterintuitive. There are a number of potential reasons for this finding. First, there is a possibility of measurement error, although this is less likely given that our calculated methods were derived using an algorithm that was verified using comparisons with invasive analyses (2). Furthermore, our reproducibility data suggest acceptable, low CVs. Second, as revealed in patients with HF with preserved EF, normal Eessb can exist despite abnormal systolic and diastolic function (10,14,22). Third, and most importantly, Ees alone is not a comprehensive measure of contractile function. Fully describing the contractile state of the ventricle requires specification of both the slope (Ees) and the intercept (V0) of the ESPVR, and in fact we detected a strong association between V0 and prognosis. In this sense, our data indicate a strong association between the contractile function of the heart and adverse outcome. However, we also note the V0 and EDV are highly correlated (R = 0.73) (Online Fig. 1), suggesting that it may be impossible to completely separate and independently measure chamber remodeling and contractile dysfunction in the intact human heart. Finally, it may also be that in our population, one of chronic HF on optimal medical therapy, VA coupling and the interaction between the ventricular and vascular systems is more important than the LV Eessb alone. Overall, the absence of the relationship between Eessb and adverse outcomes suggests that assessing chamber function via Ees alone is insufficient to gauge prognosis in human HF.
Previous studies have used precise, invasive measurements or multigated acquisition–derived estimations to demonstrate Ea and Eessb mismatch in small samples sizes of systolic HF, but were insufficiently powered to test relationships with clinical outcome (17,23). Our study corroborates these findings of Ea and Eessb mismatch in a larger population and also reveals that Ea and Ea/Eessb are strong predictors of adverse clinical outcomes. Similarly, the importance of arterial load in HF demonstrated here indicates the need for better characterization of arterial properties and their impact on outcomes in this population. Noninvasive characterization of aortic pressure-flow relationships are feasible and should yield further insight into the role of arterial load in HF in future research.
Echocardiography is inherently limited by image quality and observer variability. In our study, however, all echocardiograms were comprehensively analyzed in our core lab by a sonographer with 30 years of experience, <10% of echocardiograms were unanalyzable, and CVs were low. The methods to estimate parameters such as Eessb were previously validated by comparing Eessb with invasively measured Ees from pressure-volume relationships in humans, but still represent noninvasive estimations of Ees. Although our work provides important insight into myocardial mechanics during systole, we did not assess diastolic ventricular stiffness. It is also possible that our estimations could be improved by tonometry-derived and 3-dimensional echocardiography–derived measures (24), which are the subject of future work. We also did not have measures of myocardial contractility such as midwall fractional shortening, which may have provided additional insight into the lack of association between Eessb and outcomes.
There are important potential implications of our findings. The coupling ratio Ea/Eessb is an easily derived physiological measure that may be used to assess prognosis and response to therapy. For example, Ea/Eessb could be used as a measure to gain insight into the changes that occur with ventricular and vascular function with novel pharmacological or device therapies or with cardiotoxic agents. We may also gain insight into whether longitudinal changes over time affect prognosis. Large-scale assessment of VA coupling in population-based studies and clinical trials has the potential to provide us with an improved conceptual framework for understanding HF.
Noninvasively derived measures of ventricular vascular mechanics can be easily derived from a routine echocardiogram in patients with marked systolic dysfunction. These measures are altered in HF with reduced EF, and the extent of LV remodeling and ventricular-vascular uncoupling is strongly associated with adverse clinical outcomes. With further study, these measures can be used to gain better insight into the effects of interventions on cardiac function, suggesting that EDV, V0, and Ea/Eessb are potential therapeutic targets.
For a supplemental figure, please see the online version of this article.
Dr. Ky was supported by the NIH/Clinical and Translational Science Award KL1 RR024132, NIH K23 HL095661-01, and the Heart Failure Society of America Research Fellowship Award. This work was also supported by NIH HL088577 (Dr. Cappola). All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- coefficient of variation
- arterial elastance
- ventricular-arterial coupling, single beat
- end-diastolic volume
- end-systolic elastance
- end-systolic elastance, single beat
- ejection fraction
- end-systolic pressure volume
- end-systolic pressure-volume relationship
- end-systolic volume
- heart failure
- hazard ratio
- interquartile range
- left ventricular
- left ventricular outflow tract
- N-terminal pro–B-type natriuretic peptide
- New York Heart Association
- stroke volume
- ventricular volume at an end-systolic pressure of 0 mm Hg
- ventricular volume at an end-systolic pressure of 100 mm Hg
- ventricular assist device
- Received October 29, 2012.
- Revision received February 23, 2013.
- Accepted March 13, 2013.
- American College of Cardiology Foundation
- Solomon S.D.,
- Anavekar N.,
- Skali H.,
- et al.
- Chen C.H.,
- Fetics B.,
- Nevo E.,
- et al.
- Ky B.,
- Kimmel S.E.,
- Safa R.N.,
- et al.
- Cappola T.P.,
- Matkovich S.J.,
- Wang W.,
- et al.
- Lam C.S.,
- Roger V.L.,
- Rodeheffer R.J.,
- et al.
- Ky B.,
- French B.,
- McCloskey K.,
- et al.
- Rondeau V.,
- Mathoulin-Pelissier S.,
- Jacqmin-Gadda H.,
- Brouste V.,
- Soubeyran P.
- Rondeau V.,
- Mazroui Y.,
- Gonzalez J.R.
- Redfield M.M.,
- Jacobsen S.J.,
- Borlaug B.A.,
- Rodeheffer R.J.,
- Kass D.A.
- Chantler P.D.,
- Lakatta E.G.,
- Najjar S.S.
- De Tombe P.P.,
- Jones S.,
- Burkhoff D.,
- Hunter W.C.,
- Kass D.A.
- Asanoi H.,
- Sasayama S.,
- Kameyama T.
- Kameyama T.,
- Asanoi H.,
- Ishizaka S.,
- Sasayama S.
- Feldman M.D.,
- Pak P.H.,
- Wu C.C.,
- et al.
- Schwartzenberg S.,
- Redfield M.M.,
- From A.M.,
- Sorajja P.,
- Nishimura R.A.,
- Borlaug B.A.
- Borlaug B.A.,
- Lam C.S.,
- Roger V.L.,
- Rodeheffer R.J.,
- Redfield M.M.
- Gayat E.,
- Mor-Avi V.,
- Weinert L.,
- Yodwut C.,
- Lang R.M.